Structure-based drug design aims at identifying small molecules that can modulate protein function on the basis of their chemical complementarity with binding sites on the protein's surface. In many cases, multiple structures of a protein-ligand complex are available. These ensembles of protein structures contain information on conformational flexibility of the protein that could be useful in molecular docking studies. The DOCK suite of programs generates possible ligand orientations by matching the shape of small molecules with that of the protein surface. Scoring of the ligands orientations that are obtained is done using a grid representation of the AMBER force field energy. When docking ligands to multiple protein structures a single scoring grid, reflecting the variability of the entire ensemble, would be computationally preferable. I have developed software which allows for the derivation of such a single scoring grid from ensembles of structures by dividing the coordinates in sets of flexible and rigid atoms based on their positional standard deviation. I have applied this methodology successfully to a number of ensembles of protein-ligand complexes derived by either NMR spectroscopy or X-ray crystallography. In addition, the method allows for the selective smoothing of repulsive potentials in protein binding sites, which was shown to be useful in docking covalently binding HIV protease inhibitors to the enzyme. I am currently investigating how permissive to large ligands and computationally efficient the ensemble-based scoring grids are. Since the project involves inspection and manipulation of ensembles of protein structures and their ligands, as well as force field grids, the resources of the Computer Graphics Lab are extremely valuable for 3D visualization.